Turn a rough task into a prompt that works
Most people type one vague sentence into an AI chat and get one vague answer back. PromptPolish takes the task you already know how to describe, adds the structure a good prompt needs (a role, the audience, the tone, the output format, and checks that fit the job), and hands you a single prompt ready to paste into ChatGPT, Claude or Gemini.
Private by design: your prompt is assembled in your browser from built-in templates. Nothing you type here is sent anywhere. The only server calls this page makes are a daily counter and, if you choose to unlock more generations, your email.
Build your prompt
Four quick picks and one box in your own words. Editing is always free; only finished generations count toward the daily free limit.
Two free generations per day on this connection; after that, an email unlocks unlimited generations for the rest of the day.
Your prompt
How it works
- Pick your role and the kind of task. Ten task families cover the work people actually ask AI for: emails, reports, analysis, meetings, posts, plans, explanations, customer replies, hiring and brainstorming.
- Describe the job in plain words. No prompt jargon needed. Add the deadline, the names or the numbers in the extra details box if they matter.
- Copy the assembled prompt into your AI chat. It carries your words plus the structure: role line, audience, tone, output format, and quality constraints written for that family of task.
Why structure helps
An AI chat has no idea who you are, who the answer is for, or what good looks like for this task, unless the prompt says so. A role line sets the level and the vocabulary. Audience and tone stop the most common failure, a technically fine answer written for the wrong reader. Format instructions stop the second most common one, the right content in a shape you then have to rebuild by hand.
The constraints matter most. Each family here carries checks written for that kind of work: an email prompt demands a subject line and one clear ask, an analysis prompt orders facts before interpretation and forbids invented figures, a hiring prompt separates must-haves from nice-to-haves. They will not turn a bad task description into a good answer, but they reliably remove the avoidable mistakes.
The ten task families
- Draft an email: subject line first, one clear ask, a specific next step.
- Write a report or summary: executive summary up top, findings as headings, estimates flagged.
- Analyze data or a document: facts before interpretation, real numbers cited, gaps named.
- Prepare meeting notes or an agenda: decisions needed, action items with owners and due dates.
- Write a social or marketing post: a first line that earns the second, written for a phone screen.
- Plan a project or workflow: phases with outputs, owners, effort, risks and fallbacks.
- Explain or teach a topic: plain language first, one example per idea, check questions at the end.
- Write a customer reply: the problem acknowledged first, one clear resolution path, no false promises.
- Job or hiring task: job-related criteria only, must-haves kept apart from nice-to-haves.
- Brainstorm ideas: ten or more ideas grouped by boldness, then the three strongest picked out.
Questions
Is it free? Yes. You get two free generations a day on your connection, and an email unlocks unlimited generations for the rest of the day. There is no paid tier and no account.
Where does what I type go? Nowhere. The prompt is assembled in your browser from templates built into this page. Your task text, details and finished prompt are never uploaded. The only things that leave the page are a daily generation counter and, if you choose to unlock, your email.
Does it work with ChatGPT, Claude or Gemini? Yes. The output is plain text written for any modern AI chat tool. Paste it in, then paste your source material (the email thread, the document, the data) right under it.
Will a polished prompt guarantee a good answer? No, and be wary of tools that promise that. Structure removes the common, avoidable failures. The facts still come from you, and you should still read the answer before you send it anywhere.